Topic: multi-band compression vs. EQ

[ethan]

I have a few sets I'm working on post production where I'd like to tighten the low end and bring up the mids. It occured to me that multi-band compression might be better than EQ. Thoughts?

[Gutbucket]

Maybe.

It might be best to adress the two issues seperately, but that's not always true.. I found a nice full band compresison setting the other day for something which did both things nicely.  Regardless, you're likely to get the best results with a combination of moderate amounts of both EQ and compression (maybe different compression settings for what you want from the bass and mids), rather than trying to just use one or the other.
 
EQ settings are probably obvious and I usually try that first to see if it's enough alone.  For more midrange clarity and detail without harshing things out with too much EQ, try using parallel compression techniques that mix in a more heavily compressed signal in with the fully dynamic uncompressed signal.  That may or may not help the bass too.  For compression targeting the bottom end I find careful setting of the attack and release to get just the right amount of low end pumping key to tightening up the bass- and that's often not done as parallel.

You can do those things with two instances of regular full-band compression which might be less prone to getting off-track. Mulitband compression splits up the frequency range, runs seperate compressors on the different parts and then puts it back together again.  You can usually specify different settings for each range, but it just seems trickier to me-  a powerful tool that I haven't really had time to get a good handle on yet.

[Myco]

iZotope has a good one.

[ethan]

iZotope has a good one.

That's what I'll be using. It comes bundled with Soundforge Pro 10

[Gutbucket]

Got any pointers on good multiband settings?

I'm curious what works well for you.

[Myco]

iZotope has a good one.

That's what I'll be using. It comes bundled with Soundforge Pro 10

Yeah, that's the one I use on occassion also.

[Myco]

Got any pointers on good multiband settings?

I'm curious what works well for you.

I can't say that there is any particular setting, or group of settings that are ideal. For me it's just a trial and error sort of thing. I don't use it too often, but there have been a few occassions where I have made use of it and I've been happy with it. It has the presets that get you in a ballpark, then you can tweak individual aspects of that setting. Not an entirely scientific process for me. I bet you would understand the underlying concepts to it better than I do. I have read good reviews of it's core programming also, but it's beyond my knowledge.

[page]

(Working off of memory) On using a multi-band on the bass, I try and have the band peak and start downward about 30-40hz above where the meat of the note/kick/etc is that I'm trying to fatten. I think that makes the intersection about another 50hz or so above that and I find that works well enough. I've never tried to use the multi-band comp on anything else as EQ does the trick there (or it requires other effort that a compressor just won't fix).

I personally like how Ozone's comp lets you target via either RMS or Peak based. What I don't like about it is the only lookahead function is in the mucho-expensivo edition (which is why I only use it on the bass where I think any pumping that occurs sucks less than if it occurs elsewhere).

[Matt Quinn]

Maybe.

It might be best to adress the two issues seperately, but that's not always true.

This would 100% be my take on it, and maybe it's just because I use Ableton, which has a ridiculously awesome amount of flexibility when it comes to building your own serial/parallel device chains. They are 2 separate tools for 2 separate issues, for me it doesn't make sense to try to kill 2 birds with one stone when you've got 2 perfectly good stones.

In general, EQ comes before compression. I'd get the tonal balance as close as possible to what you want before moving on to dynamics.

[page]

In general, EQ comes before compression. I'd get the tonal balance as close as possible to what you want before moving on to dynamics.

Do you find that your compression alters the tonal balance? I've been hit and miss, sometimes I find it brightens the source (e.g. cymbals are sitting just under the loud stuff which gets tramped down thus revealing the cymbal more), and sometimes I don't notice it.

So I take the opposite approach; comp then EQ. I don't think there is necessarily a good/bad approach in that regard as long as it works.

[Gutbucket]

Myco, my approach is similar in that I know a bit of what is going on, or how it is supposed to work at least, but for the most part I just use that as a starting point and home things in from there by ear.  Lots of fiddling and trial & error.

Page, are you using RMS or peak detection for that bass comp?

The thing about pumping is that when using compression to tighten up the bass, I'm sort of using that effect to add some missing dynamic to an otherwise flabby and ill-defined bloated bass or kick after getting the overall level general right with EQ.  Maybe I'm using the term pumping incorrectly and bouncing is a better description for what I'm shooting for here, where by contrast, pumping is more about the unwanted side-effect on the other frequency ranges outside the bass range being targeted (possibly appropriate as an effect for mixers but not what we want as a more or less transparent mastering tool)?

That's in contrast to the standard application of compression in reducing dynamics, and more of an odd cookie for me to get set correctly.  I have a harder time getting it to work, and sometimes it doesn't at all, but as mentioned the attack and release settings working closely in rhythm with the music are key.  Seems a game of letting whatever exists of the initial transients through before the compressor clamps down and lowers the level, then releasing smoothly before the next event to let it through uncompressed again.  Possibly a good job for the bottom range of a multiband comp since the 'bounce' would be limited to that band alone and not 'pump' the other bands.

Matt, interesting to hear about Ableton (never used it).  The device chain building aspect appeals to my way of thinking and I've always wanted an easier way to do that or at least visualize the signal flow graphically.

I can see arguments for EQ before or after comp.  I sometimes do both, depending.. usually EQ before compression at the track level and afterwards at the master bus, simply because they fall in that order without rearranging things, and I'm usually working iteratively and addressing one aspect at a time then moving on to something else.  But that's more of a practical workflow thing.  Yet consider electric guitar rigs and processors.  Totally different goal since it's all about imparting color, but where the EQ goes in the signal chain makes a big difference in sound and there is often an EQ stage both before and after distortion stages for instance.

It might be that controlling major frequency and level issues with one stage of EQ and dynamics before going on to fine tuning the more balanced result of that with other, more targeted EQ and/or compression settings is called for.  Multiple small adjustments often seem to work better for me than one big adjustment.

[page]

Page, are you using RMS or peak detection for that bass comp?

I favor peak detection but it's more of a "what do I want to accomplish" (and what am I better at setting up quickly due to experience). I'm trying to contain spikes, and then second, round out a tone more then anything. I've actually daisy-chained compressors for that reason. Use something wtih lookahead and adjustable frequency detectors for the main sections (e.g. 300-8000) and then come back around with Ozone's in multi-band mode for just that sub 300hz level.

Maybe I'm using the term pumping incorrectly and bouncing is a better description for what I'm shooting for here, where by contrast, pumping is more about the unwanted side-effect on the other frequency ranges outside the bass range being targeted (possibly appropriate as an effect for mixers but not what we want as a more or less transparent mastering tool)?

True, I guess it's a "whats the problem at hand". I favor a fast attack on my comp by default because I don't have the luxury of individual micing so I'm battling wild dynamics that are coming from a range of instruments. That's why I have a healthy penchant for lookahead functionality. Softube's FET Compressor has one and the color it imparts works with with my soft mics. If I was using something more sharp or bright, I'd probably look elsewhere.

It might be that controlling major frequency and level issues with one stage of EQ and dynamics before going on to fine tuning the more balanced result of that with other, more targeted EQ and/or compression settings is called for.  Multiple small adjustments often seem to work better for me than one big adjustment.

Mildly off topic:
I was playing with a plugin recently that had an analog distortion modeler with it and I found if I knocked the saturation down to almost nothing (as recommended elsewhere) that it was pleasant on the master bus. The real benefit though was that coupled with the wet setting. In this particular case, the gear that was being emulated had a punch or compress effect that was just enough to benefit the mix without me having to dicker with it elsewhere (and being able to adjust just how much punch/flatening to introduce). Used sparingly it was really nice, but I wouldn't crank the thing by any means. Right tool for the right job.

[ethan]

thanks all I'm finding the discussion quite interesting....can't wait to get off work and get to my DAW

[Gutbucket]

Yep, good discussion.  Thanks for the thread, E.

Page, that makes sense.  Peak detection should give a faster response and wider range of input values which is probably better suited to this stuff.. RMS a slower, more averaged response, probably making RMS more appropriate for envelope-like signal leveling over longer time periods.  Yet I've never gotten that to work well sound quality-wise and typically resort to manually dawing volume envelopes for longer term level adjustments.  Of course that could be the tools I'm using or the way I'm using them.

When I use a compressor alone to bring up details in the midrange and top (not paralleled) the attack/release setting is what seems to control color somewhat by how much transient edge makes it through to brighten things up or tone it down.  Maybe that's something of a corellary to getting a bit of dynamic bounce in an otherwise loose bass with somewhat longer time settings.  I never really thought of it that way before because it seems like an entirely different effect subjectively, with not only different time settings but also less threshold and make-up gain, and often set more sparingly as you mention.  It also might have something to do with 'tuning' compressor time settings appropriately for the mid and treble sounds being somewhat more general in nature than tuning for bottom, as snare and cymbal hits are typically more universal in their faster envelope than rhythmic bass and kick interaction.

[F.O.Bean]

I NEVER EQ. ONLY use the Multi-Band Compressor in WaveLab and I LOVE IT you can easily screw up a recording when EQing, but MBC is much easier to deal with

[page]

When I use a compressor alone to bring up details in the midrange and top (not paralleled) the attack/release setting is what seems to control color somewhat by how much transient edge makes it through to brighten things up or tone it down.  Maybe that's something of a corellary to getting a bit of dynamic bounce in an otherwise loose bass with somewhat longer time settings.  I never really thought of it that way before because it seems like an entirely different effect subjectively, with not only different time settings but also less threshold and make-up gain, and often set more sparingly as you mention.  It also might have something to do with 'tuning' compressor time settings appropriately for the mid and treble sounds being somewhat more general in nature than tuning for bottom, as snare and cymbal hits are typically more universal in their faster envelope than rhythmic bass and kick interaction.

On the bottom end, it's generally more forgiving. Creating rumbly goo while not optimum, can be acceptable in a "color" type way where as creating a gooey mess in the mid-upper registers sort of sucks. Stylistically one works better than the other even though the net effect is the same.

As far as tightening bass (and adding a touch of color), have you considered phase cancelation during coloration such as via an enhancer/exciter? For example in Ozone, I can select a type of excitement, and the band delay before it feeds it back in thus creating a sort of comb filter. Figure out the correct settings and you can tighten stuff up really well. It's not my favorite approach and I don't use it for that purpose enough to get good at it, but it's quite possible. On the hardware side, the sonosax sx-m2 (among others) can do that as well so it's not a pure synthetic effect. On that note, in Ozone (unlike reality) you can have pre-echo instead of just post-echo with negative delay numbers. Try working in 0.1ms increments (in both directions) to see what it does one day. The same can be said about the stereo spreader. Set it to about 15% for learning purposes and then fiddle with the delay and watch the stereo image swing around, especially in headphones and then again with nearfields (it won't always move the same way depending on what you use). Fun stuff to goof off with, but it only has limited applicability for what we typically see.

So back on topic:

I think the big thing I've found with compression is tuning the detector so that
A) It compresses what I want in a manner that I think is predictable.
B) It doesn't trigger when I'm not asking it to due to material outside of my target range.
C) In the process of doing A, it doesn't trash the material that is present in B.

As an example:

    sample-post_prod.16bit.flac
    http://fyels.com/8Rn - 38.04 MB

    sample-original.24bit.flac
    http://fyels.com/HRn - 56.68 MB

(compression, among other things used in the sample):

a fast attack FET-style compressor (from Softube)
- ~0.05ms attack, ~7ms release, 2ms lookahead
- detector adjustments: LPF at 8khz, HPF at 300hz
- parallel comp/injection: 25%, maybe 15%.
- compression ratio: 2.5:1

Kick drum compression (Izotope):
- 0.1ms, 20ms release, 3:1, soft knee.
- 80% wet mix under 200hz, 60% (and dropping) at 250hz. See screenshot below

Yes it's an in your face mix but thats partially the nature of the band and compromises for a consistent effect throughout the show. It's not flawless by any means, but it's an example of how you can use 2 compressors, even one that isn't specifically multi-band, in such a way that they compliment each other (and it's the only example I had handy). Had I used just the first one, it would fluctuate more and I ended up trying to EQ things only to create holes when the bottom end wasn't prevalent. It needed to be fattened via compression more so than EQ. (To truly evaluate them, knock the 16 bit sample down by 15.3db, that should match the RMS of the left channels.)

As always, ymmv, that's just what I get out of it.

edit: Now that it's not the late evening or early morning, I realize it would probably be beneficial to have an intermediary (third) sample with one compressor and not the other... If I remember that tonight I'll try and dig it out and turn off that second comp.

[Gutbucket]

Thanks for the detailed write up on your approach. I’ll try and make some time to play around with what you are doing. Mental note to check out a trial of Ozone again, it's been 4 or 5 years or so.

It's been somewhat more difficult for me to get the bottom end doing what I want with compression, probably because I haven't put as much time into that as compressing for level control and compressing to bring out detail, texture and EQ-like tonal control in the midrange and lower treble.  Compressing for detail and texture I can get to work much better, where compressing for level control works but is more frustrating in trying to keep the quality of sound I want, so on important stuff I end up spending lots of time drawing volume envelopes instead of compressing for level, and even searching out and redrawing each wild peak instead of using a limiter which always seems to kill a bit of the liveness I love.


Looking at all this from the big perspective-
When I sit back and think about mastering my recordings, I think about what I want to do, and how I’d like to get there. I then need to adapt that to the tools I have and the ways I can work.  That’s usually the frustrating part.  I know what I want, and the technical aspects of it, but getting there isn’t as easy as it seems it should be.

The primary things I want to adjust in raw recordings are:

Overall frequency balance
Overall dynamics (long term- the entire show)
Short term dynamics (controlling wild peaks)
Timbre and detail
Correction of general problems (ill defined bass, instrument balances, etc)
Correction of specific short term problems (editing noises, pops, claps, coughs, clips, etc)

And that’s the order I think of them, but when I analyze it, I realize I should actually address it in almost the reverse order:

Fix the short term problems (peaks, clips, pops, etc)
Correct the short term EQ that varies with dynamics
Correct the overall dynamics
Correct the general problems
Timbre and detail
Overall EQ

Part of what makes things difficult is that EQ and dynamics are so interdependent.  I want different EQ settings based on the level of the music. I often find myself wishing for an EQ which tracks dynamics (longer term than RMS, more like a program leveler in it’s time response) and allows me to set different EQ curves for various signal levels which it smoothly morphs between as the overall level changes.

I’d want several curves at various loudness levels, since the difference between the EQ I want for the non musical portions (between song banter, crowd noise, etc) and the musical portions is often greater than the difference in EQ I’d like to make with level changes within the music.

The short term dynamic problems like peaks and pops should be fixed before that dynamic EQ stage so they don’t effect the level sensing of the EQ, but the overall dynamics should probably be adjusted afterwards as the overall level difference will end up being reduced and therefore harder to ‘sense’ by the dynamic EQ algorithm.

Know of any plugin that does this or a way of setting up a chain to do this?  I think it’s a good opportunity for a VST designer!

Samplitude does have EQ automation which I should look into, but I think it's more like manually drawing volume envelopes. 

[Matt Quinn]

Part of what makes things difficult is that EQ and dynamics are so interdependent.  I want different EQ settings based on the level of the music. I often find myself wishing for an EQ which tracks dynamics (longer term than RMS, more like a program leveler in it’s time response) and allows me to set different EQ curves for various signal levels which it smoothly morphs between as the overall level changes.

I’d want several curves at various loudness levels, since the difference between the EQ I want for the non musical portions (between song banter, crowd noise, etc) and the musical portions is often greater than the difference in EQ I’d like to make with level changes within the music.

I'm going to really recommend you check out Live. I know that learning a new DAW can be a huge headache, but I suspect you will love it. Specifically in regards to the question about, you could certainly achieve something very close to that with the use of Racks, Chains, and Zones.


In Live, you can combine devices - any devices, Live's internal instruments/effects etc, VST, or other racks - into neat little packages called Racks. They look like this (right side):




Within each rack, you can create as many Chains as you like. Chains are parallel processing systems.

Rack with it's chains showing:





So for example, if you planned on using 4 main different EQ curves for a set, set up 4 chains, place one of the EQ devices at the end of each. Each chain has it's own volume, send, & pan controls, as well as a kill switch. You have a lot of options as to how you decide what signal goes to which processing chain, mainly decided by what Live calls 'Zones'. There are zones for key zone, velocity, and chain select, but for audio you would be mainly working with chain select.

Zones:




Zones allow you to define ranges under which signals will be allowed to flow to the chains. IE if you have your 4 different EQ curves set up, you can split your zones into 4 separate parts, and then use the chain selection to automate at what point the signal starts flowing into each chain. You can overlap chains as well, and fade between them as the signal passes through the crossing.



If all this is too complicated, you can simply automate any parameter you'd like in Live in a straightforward way.



Each track has a drop down containing all the devices used in the track. Simply select the parameter you'd like to automate, and a new lane appears below your audio. You can use a pencil to draw automation, or lines & breakpoints to make smooth transitions. You can also automate simply switching between presets.

Whew, that was a mouthful! If I was unclear, feel free to ask about anythnig I mentioned!

[Gutbucket]

Thanks Matt.  Does look interesting, especially in arranging signal paths and device chains.  From what you describe I don't immediatley see a way to have it automatically switch between four different EQ chains based on signal level, but that may be possible.  I'd like to look deeper when I get a chance.

[Matt Quinn]

Thanks Matt.  Does look interesting, especially in arranging signal paths and device chains.  From what you describe I don't immediatley see a way to have it automatically switch between four different EQ chains based on signal level, but that may be possible.  I'd like to look deeper when I get a chance.

I posed your question on the Ableton Forum, and someone suggested using pairs of Gate (internal Live effect) to define the audible ranges to allow the signal to pass through. Gate has both a normal gate function (cutting off low level signals) and a 'Flip' switch that causes it to work in reverse, ie cutting off signals *above* a certain threshold.


Actually, now that I'm looking at it in the manual, the Gate also has a sidechain function, which allows you to trigger it from any audio source. You'd probably have to play with some routing to get it right, but I see no reason why you could not just duplicate the wav you are working on to a second track, and use that to feed the sidechain on the gate.

[ethan]

Well after about 10 hours of playing the last couple nights with my Jam Cruise matrix recordings I can say I really like using multi-band compression. I'm still not quite confident about attack/release but I managed to get what I was looking for on several of my recordings which just needed some "touch-up".

To my ears compressing low end makes more space for everything else and opens things up.

[Gutbucket]

I played with the trial version of Voxengo Soniformer a few years ago and really liked it.  I should revisit that.  It's sort of an uber multi-band compressor tool, aimed at mastering to keep the spectrum in control, with very high band count and provision for drawing varying threshold and time curves across all bands quite easily.  Check it out if curious.  As I recall it was easier than others I tried to get good sounding, transparent results.

[morst]

I NEVER EQ. ONLY use the Multi-Band Compressor in WaveLab and I LOVE IT you can easily screw up a recording when EQing, but MBC is much easier to deal with

Same here. EQ causes phase shifting and comb filtering.

Well after about 10 hours of playing the last couple nights with my Jam Cruise matrix recordings I can say I really like using multi-band compression. I'm still not quite confident about attack/release but I managed to get what I was looking for on several of my recordings which just needed some "touch-up".

To my ears compressing low end makes more space for everything else and opens things up.

your ears are not making stuff up, it can be very true. Keep twiddling the settings on attack and release. Set 'em way too fast or way too slow and listen to hear what they're doing. Then tune in on your best settings.

[Matt Quinn]

I NEVER EQ. ONLY use the Multi-Band Compressor in WaveLab and I LOVE IT you can easily screw up a recording when EQing, but MBC is much easier to deal with

Same here. EQ causes phase shifting and comb filtering.

Not if you use a Linear Phase Equalizer. I use this one.

http://www.ikmultimedia.com/t-racks/moreinfo/moreinfo6.php


Also, I do not agree with the idea that MBC is somehow idiot proof. You can 'screw up' a recording with MBC just as much as you could with EQ/Compression. It's about how you use your tools, not what tools you have.

[page]

Also, I do not agree with the idea that MBC is somehow idiot proof. You can 'screw up' a recording with MBC just as much as you could with EQ/Compression. It's about how you use your tools, not what tools you have.

Agree with all of that. The same arguement can be said about getting a great recording with midtier mics versus a meh recording with awesome equipment.